Marshallese phonology

This article needs additional citations for verification. Please help improve this article by adding citations to reliable sources. Unsourced material may be challenged and removed. (February 2011)

This article may contain original research. Please improve it by verifying the claims made and adding references. Statements consisting only of original research may be removed. More details may be available on the talk page. (April 2012)

Marshallese Alphabet, Public Library, Delap–Uliga–Djarrit (DUD)

Marshallese phonology is characterized by:

• 22 consonant phonemes with contrasting primary and secondary places of articulation.
• 4 vowel phonemes in a vertical vowel system.
• A large inventory of allophones.

Consonants

Marshallese has 22 consonant phonemes, each written with between zero and two letters.

• Marshallese has a number of consonants with contrasting secondary articulations:
  • palatalized consonants
  • velarized consonants
  • labio-velarized consonants
• Consonant voicing and devoicing do not phonemically contrast in Marshallese, and there may be free variation in articulation between voiced and voiceless consonants.^[1] Marshallese obstruent phonemes are traditionally described using voiceless IPA symbols /pʲ pˠ tʲ tˠ k kʷ/. In practice, obstruents are voiceless /p p̴ t͡ɕ t̴ k kʷ/ at the beginning and end of a word and when the obstruent is geminated, but are voiced /b b̴ ʑ d̴ ɡ ɡʷ/ between two vowels.^[2] (Compare Korean phonology, where something similar happens.)
• The palatalized coronal obstruent /tʲ/ is sometimes described^[3] as a palatal obstruent /c/, because there is no palatalized dorsal obstruent phoneme to contrast against.^[1] The pronunciation is actually usually an alveolo-palatal articulation of either [t͡ɕ] or [ɕ] when voiceless, or [ʑ] or [d͡ʑ] when voiced.^[2] While these articulations may vary between people and dialects, generally speaking, the voiceless affricates are more common at the edges of words, and the voiced fricatives are more common word-internally (this article uses [t͡ɕ] and [ʑ] for this reason).[d͡ʑ] when voiced.^[2] These are all allophones in Marshallese, and foreign postalveolar consonants and alveolar sonorants are typically assimilated as Marshallese ⟨j⟩.^[2]
• There is no labio-velarized coronal obstruent phoneme.^[1]
• The rhotic consonants /rʲ rˠ rˠʷ/ are all trilled.^[2]
• The dorsal consonants /ŋ ŋʷ k kʷ/ are pronounced with a more retracted tongue than in English.^[2]
• The plain approximants /ɦʲ ɦˠ ɦˠʷ/ are purely horizontal approximants, and are not strictly semivowels [j ɰ w] with limited vowel height as might be expected of approximants in English or French.^{[4][disputed – discuss]} With undefined vowel height, plain approximants audibly all but vanish in proximity to the vowels they neighbor,^[1][4] but as consonants are still phonemically distinct in how they color those neighboring vowels' backness and roundedness.^[1][3] This is related to why plain approximants often go unwritten in orthography. Marshallese in fact does have semivowels, but when articulated they are a cooperation of the two-dimensional contrasts of backness/roundedness-defined approximants and height-defined vowels, without being independent consonant phonemes on their own.

Vowels

Marshallese has a vertical vowel system of just four vowel phonemes each with several allophones:

Marshallese vowels are not specified along the front-back and rounded-unrounded dimensions, but on the height and ATR dimensions (see the IPA classification of vowels in the table on the right). This means that a given vowel phoneme will have several different phonetic realizations.

For example, the high vowel phoneme /ɨ/ may alternately be pronounced [i], [ɯ], [u], [i͡ɯ], [i͡u], [ɯ͡i], [ɯ͡u], [u͡i], [u͡ɯ], depending on the context:^{[citation needed]}

• Specifically, vowels next to palatized consonants become front unrounded [i], [e], [ɛ], [æ],
• vowels next to velarized consonants become back unrounded [ɯ], [ɤ], [ʌ], [ɑ],
• and vowels next to labialized consonants become back rounded [u], [o], [ɔ], [ɒ].
• When between two consonants of different types (e.g., a velarized consonant and a labialized consonant), the vowels become diphthongs, beginning with the surface form found next to the preceding consonant, and ending with the surface form found next to the following consonant (e.g., in the case of a vowel between a velarized and labialized consonant, the diphthongs would be [ɯ͡u, ɤ͡o, ʌ͡ɔ, ɑ͡ɒ]:

Syllable and phonotactics

This section's factual accuracy is disputed. Please help to ensure that disputed facts are reliably sourced. See the relevant discussion on the talk page.

Syllables in Marshallese follow CV, CVC, and VC patterns.^{[1][page needed]}

All syllables begin and end with a consonant, with a vowel at the syllable's nucleus.^{[1][page needed]}

• The smallest possible one-syllable word is structured /CVC/.^{[1][page needed]}
• Multiple /CVC/ sequences can form a multi-syllable word, meaning that consonant clusters of no greater than two consonants can exist inside a word if surrounded by vowels: /CVCCVC/.^{[citation needed]} The two consonants in a cluster are not required to agree by palatalization, velarization or labialization.^{[citation needed]}
• It is also possible for a syllable's final consonant and the following syllable's first consonant to be shared, so /CVCVC/ is also possible.^{[citation needed]}

In all cases, the vowel of each syllable assimilates to the consonants at both its sides:^{[citation needed]}

• If both consonants are palatalized, the vowel becomes a front vowel.
• If both consonants are velarized, the vowel becomes a back vowel.
• If both consonants are labialized, the vowel becomes a rounded back vowel.
• If the two consonants are of two different categories, the vowel forms a diphthong in between them.

These multiple vowel articulations are a product of the different consonant phonemes that neighbor them on each side. They are not vowel phonemes in their own right, as Marshallese vowels are phonemically distinguished only by vowel height. Therefore, a vowel's backness or roundedness is relevant to the consonant phonemes, and not to the vowel phoneme.

Marshallese orthography permits that words can begin and end with vowels, but this merely means the consonant phoneme on the "vowel" end of the word is actually an unwritten approximant:^{[citation needed]}

• /ɦʲ/ if the vowel is a front vowel.
• /ɦˠ/ if the vowel is a back vowel.
• /ɦˠʷ/ if the vowel is a rounded vowel.

Semivowels

Any Marshallese vowel can also have an asyllabic semivowel allophone.^{[citation needed]} This can happen when a vowel is separated from a more dominant vowel

This article may be too technical for most readers to understand. Please help improve this article to make it understandable to non-experts, without removing the technical details. The talk page may contain suggestions. (May 2012)

by a single plain approximant consonant /ɦʲ ɦˠ ɦˠʷ/.^{[citation needed]} The vowel allophones [i ɯ u] have strong semivowel allophones [j ɰ w], which are approximant allophones stronger and more complex than the heightless approximant phonemes /ɦʲ ɦˠ ɦˠʷ/. These strong semivowels have specific vowel height more similar to approximants in English or French. But even more open vowels can also have approximant allophones that contrast with the close semivowels, which can also be seen in languages like Romanian. Even diphthongs (which are themselves already smooth-transitioning allophones) can have semivowel allophones.

Some examples:

• io̧kwe^[5]: /ɦʲɨ̯ɦʲakʷɜɦʲ/ or [jæ͡ɒɡʷɔ͡ɛɛ̯]
• jaab^[6]: /tʲaɦˠapˠ/ or [t͡ɕæ͡ɑʕɑp̴]
• jouj^[7]: /tʲɘɦˠʷɨtʲ/ or [t͡ɕe͡owu͡it͡ɕ]
• tāākji^[8]: /tˠaɦʲaktʲɨɦʲ/ or [t̴ɑ͡ææ̯æ͡ɑɡʌ͡ɛʑij]

In situations where a semivowel has the same vowel height as the syllabic vowel it neighbors (separated only by the plain approximant), for instance in the word tāākji, the fundamental consonant and vowel phonemes can usually still be told apart, particularly when the plain approximant between the vowels differs in its secondary place of articulation from the other two closest consonants.^{[citation needed]} But if all three consonant phonemes agree in secondary place of articulation, even native speakers can struggle to tell them apart if their difference is not clearly enunciated.^{[citation needed]} For example:

• jān^[9]: /tʲanʲ/ or [t͡ɕæn]
• jāān^[10]: /tʲaɦʲanʲ/ or [t͡ɕææ̯æn]

Consonant clusters

Though any arbitrary consonant clusters can be written in Marshallese orthography, only certain consonant pairs can form stable clusters in Marshallese.^{[1][page needed]} Any cluster that includes a plain approximant consonant /ɦʲ ɦˠ ɦˠʷ/ is stable. Pure geminates (two of the same consonant) are also stable. Besides those, there are also:

• Nasal-Obstruent: /mʲpʲ mˠpˠ nʲtʲ nˠtˠ ŋk ŋʷkʷ/
• Nasal-Liquid: /nʲrʲ nˠrˠ nˠʷrˠʷ nʲlʲ nˠlˠ nˠʷlˠʷ/
• Liquid-Liquid: /rʲlʲ rˠlˠ rˠʷlˠʷ lʲrʲ lˠrˠ lˠʷrˠʷ/
• /lʲtˠ/ and /lˠtˠ/

Of these, all clusters agree by both their primary and secondary places of articulation (but not always in manner of articulation), with the exception of /lʲtˠ/.^{[citation needed]}

Assimilation

Certain other consonant pairs assimilate to form stable clusters or geminates, even across word boundaries.

When both consonants are bilabial:^{[citation needed]}

When both consonants are coronal (empty areas are epenthesis):^{[citation needed]}

When both consonants are velar:^{[citation needed]}

For example, Arņo^[11] /ɦˠarˠnˠɜɦˠʷ/ is pronounced [ʕɑn̴n̴ʌ͡ɔɔ̯], as if spelt Aņņo.

Epenthesis

Other consonant pairs cannot combine into a stable clusters, so an epenthetic vowel is inserted between them.^{[citation needed]} The height of this epenthetic vowel is transitional between the two nearest vowels.^{[citation needed]} Many other westernized Marshallese place names spell out the epenthetic vowels, including:

• Ebeye, from earlier Ebeje, from Epjā^[12]: /ɦʲɜpʲ​​tʲaɦʲ/ or [ɛ̯ɛb​ɛ​ʑææ̯], as if spelt Epejā
• Erikub, from Ādkup^[13]: /ɦʲarʲ​​kʷɨpʲ/ or [æ̯ær̟​ɛ͡ɔ​ɡʷu͡ip], as if spelt Ādōkup
• Kwajalein, from Kuwajleen^[14]: /kʷɨɦˠʷatʲ​​lʲɜɦʲɜnʲ/ or [kʷuwɒ͡æʑ​æ​lɛɛ̯ɛn], as if spelt Kuwajāleen
• Majuro, from Mājro^[15]: /mʲatʲ​​rˠɘɦˠʷ/ or [mæʑ​ɛ͡ʌ​r̴ɤ͡oo̯], as if spelt Mājōro
• Namorik, from Nam̧dik^[16]: /nʲamˠ​​rʲɨk/ or [næ͡ɑm̴​ʌ͡ɛ​r̟i͡ɯk], as if spelt Nam̧ōdik
• Rongelap, from Ron̄ļap^[17]: /rˠʷɜŋʷ​​lˠapʲ/ or [r̴ʷɔŋʷ​ɔ͡ʌ​ɫɑ͡æp], as if spelt Ron̄wōļap
• Rongerik, from Ron̄dik^[18]: /rˠʷɜŋʷ​​rʲɨk/ or [r̴ʷɔŋʷ​o͡e​r̟i͡ɯk], as if spelt Ron̄wōdik
• Ujelang, from Wūjlan̄^[19]: /ɦˠʷɨtʲ​​lʲaŋ/ or [wu͡iʑ​e​læ͡ɑŋ], as if spelt Wūjelan̄
• Uliga, from Wūlka^[20]: /ɦˠʷɨlʲ​​kaɦˠ/ or [wu͡il​e͡ɤ​ɡɑʕ], as if spelt Wūlōka
• Utirik, from Utrōk^[21]: /ɦˠʷɨtˠ​​rˠɘk/ or [wu͡ɯd̴​ɯ​r̴ɤk], as if spelt Utūrōk

With plain approximants

Marshallese consonant clusters can include any combination of plain approximant consonants /ɦʲ ɦˠ ɦˠʷ/,^{[citation needed]} despite these approximants being almost transparent in proximity to a vowel. When they occur in a cluster with a different kind of consonant, they result in a syllable hiatus favoring the approximant:^{[clarification needed]}

• Bokaak^[22]: /pˠɜ​kʷɦˠ​ak/ or [p̴ʌ͡ɔ​ɡʷʕ​ɑk]
• Teļap^[23]: /tˠɜ​ɦʲlˠ​apʲ/ or [t̴ʌ͡ɛ​ɛ̯ɫ​ɑ͡æp]
• M̧ajeļ^[24]: /mˠa​ɦˠtʲ​ɜlˠ/ or [m̴ɑ​ʕʑ​ɛ͡ʌɫ]

If the approximant differs in secondary place of articulation from the non-approximant it is paired with, the distinction is at least still clear because of the different horizontal locations of the neighboring vowel allophones.

When the approximant agrees in secondary place of articulation, the distinction can be more subtle:

• Ratak^[25]: /rˠa​ɦˠtˠ​ak/ or [r̴ɑ​ʕd̴​ɑk]
• Mājej^[26]: /mʲa​tʲɦʲ​ɜtʲ/ or [mæ​ʑɛ̯​ɛt͡ɕ]

Since syllables split at consonant clusters anyway, the difference in pronunciation would be small even if the approximant were not there, and the approximant needs to be articulated by a just-perceptable glide.

When both consonants in the clusters are plain approximants with different articulation, they are clear from the abrupt change in horizontal position of the neighboring vowels, but otherwise are split by hiatus:

• Ibae^[27]: /ɦʲɨpˠa​ɦˠɦʲ​ɜɦʲ/ or [ji͡ɯb̴ɑ​ʕɛ̯​ɛɛ̯]
• M̧aļoeļap^[28]: /mˠalˠɜ​ɦˠʷɦʲ​ɜlˠapʲ/ or [m̴ɑɫʌ͡ɔ​ɔ̯ɛ̯​ɛ͡ʌɫɑ͡æp]

Things become less clear when the cluster is a single type of plain approximant which is geminated:

• indeeo^[29]: /ɦʲɨnʲrʲɘ​ɦʲɦʲ​ɜɦˠʷ/ or [jinr̟e​e̯ɛ̯​ɛ͡ɔɔ̯]

Stress

This section is empty. You can help by adding to it.

References